Investigating adverse events following immunisation with pneumococcal polysaccharide vaccine using electronic General Practice data

BackgroundIn early 2011, following an increased number of reports of severe vaccine-related injection site reactions, Australian authorities recommended against administering repeat doses of the 23-valent pneumococcal polysaccharide vaccine (23vPPV) in otherwise healthy adults. The aim of this study was to assess a source of electronic medical record data from primary care providers (General Practitioners, GPs), for validity and ability to retrospectively detect this adverse event signal.MethodsThe General Practice Research Network (GPRN) holds data routinely collected from a representative sample of Australian GPs. Data were extracted on persons 18 years or older who had received at least one dose of 23vPPV or influenza vaccine (as comparator) between January 2002 and June 2012. Increases above background levels were assessed using 95% confidence intervals of reaction rates, calculated from the Poisson distribution of counts.ResultsThere was an average of 253 practices and 532 GPs contributing data per year. Over the study period there were 95,760 recorded 23vPPV administrations and 823 reactions, of which 233 were local. For influenza vaccine the numbers were 683,829 doses, 3001 and 387 respectively. Patterns of vaccinations and reactions were consistent with known safety profiles. There were 3 local reactions following 23vPPV in early 2011 (235/100,000 doses, 95% CI 49–717), which was not significantly different to the historical average (260, 225–298). We estimate that this system could have detected a 3-fold increase over background levels.ConclusionsUsing GP consultation data, we were unable to confirm an increase in local reactions detected by passive surveillance, suggesting that this apparent signal was artefactual. GP consultation data captures large numbers of vaccine recipients and medically attended adverse reactions at low cost. If available in a timely manner and expanded, this system has significant potential for use in validation of apparent signals from passive surveillance.